SPI Report for the LHCC Comprehensive Review

SPI Report for the

LHCC Comprehensive Review

Stefan Roiser

for the SPI project

19. Nov. 2007 LHCC Comprehensive Review - SPI 2

SPI services

• Software process support– Software testing, configuration & release

• Documentation support– Code documentation, web services

• Collaboration tools– Hypernews, Savannah

• Hardware infrastructure– Build servers, special purpose machines

Software


Time to user

• The time from a bug report / feature request until it is used by the physicists in the experiments– Especially important to minimize this time when

running with “live” data.


Nightly Builds

• Recommendation of the 2006 internal review• Motivation

– Early integration into experiment software– Fast feedback to LCG/AA developers of changes

• Usage– Building with own experiment nightlies on top– Developers testing their code on LCG/AA stack

• Used by– LHCb– Atlas


Nightly Builds (ctd.)

5 configurations

LCG/AA projects

13 possible platforms

Builds & Tests

test evolution


Operating System

eg. Athena, CMSSW, Gaudi

Linux (slc) 32/64 bit, Mac OSX, Windows

External Software

LCG / AA software

Experiment software

Operating System

eg. Athena, CMSSW, Gaudi

Linux (slc) 32/64 bit, Mac OSX, Windows

External Software

LCG / AA software

Experiment software

Software release

56 packages, eg. Boost, Python, Qt, etc…

COOL, CORAL, POOL, RELAX, ROOT, SEAL

LCG release

• rubberstamps all package versions• provides several products

• afs installation• web pages• src tar balls• bin tar balls• documentation


Software processes

mysql

externs/w

scramcmt install.mgr.

cron /pythontar

ballsxml /

python

python

nightlybuilds

projects/w

websites docu

xml

historicallygrown


Software processes new

externs/w

install.mgr.

tarballs

nightlybuilds

projects/w

websites docu

Python /LCGCMT


LCGCMT

• Is a CMT based project • Contains the configuration information

– How to build software packages from source• Using sh/bat scripts

– Extra information • Eg. homepage, package maintainer, description, etc.

– In addition for other CMT based projects• How to use/build against the packages• Runtime environment


Example: package build

package gsl

use LCG_BuildPolicy v* LCG_Builders

macro gsl_build_config_version "$(GSL_config_version)"set LCG_tarfilename "gsl-$(gsl_build_config_version).tar.gz"

apply_pattern buildscripts_local

#!/bin/sh

cd ${LCG_builddir}tar xvfz ${LCG_tarfilename}cd gsl-${LCG_package_config_version}./configure --prefix=${LCG_extdir}/GSL/${LCG_package_config_version} /${LCG_CMTCONFIG}

#!/bin/sh

cd ${LCG_builddir}/gsl-${LCG_package_config_version}make

#!/bin/sh

cd ${LCG_builddir}/gsl-${LCG_package_config_version}make install

lxplus:~ >setenv CMTCONFIG slc4_amd64_gcc34lxplus:~/LCGCMT/LCG_Builders/gsl/cmt >cmt pkg_getlxplus:~/LCGCMT/LCG_Builders/gsl/cmt >cmt pkg_configlxplus:~/LCGCMT/LCG_Builders/gsl/cmt >cmt pkg_makelxplus:~/LCGCMT/LCG_Builders/gsl/cmt >cmt pkg_install


Benefits of new s/w process

• Avoid duplication of information• Easier maintenance / automation• All parties use the same info (nightly builds,

release, users rebuilding the software)• Speedup time to user

Current configuration

LCG 53f has been

built with these tools

Communication


Hypernews

• Bulletin board system used by LHC experiments

• Moved to new IT provided hardware– 1 production machine, 1 hot standby

• Helping with instance maintenance – Managing user DB, setup of for a

• Atlas and CMS heavy users– Also used in Alice and Totem– LHCb is rethinking their usage


Savannah

• Savannah is the main bug reporting system in LCG/AA

• Moved to new IT provided hardware• Used by many other parties (IT, Grid, etc.)• Reasonable system / user maintenance

0

50

100

150

200

250

aug'02feb'03aug'03feb'04aug'04feb'05aug'05feb'06aug'06feb'07aug'07

Active Projects

0

500

1000

1500

2000

2500

3000


Registered Users

0

50

100

150

200

(days)


Bug Average Lifetime

05000

1000015000200002500030000


Submitted Bugs

0

500

1000

1500

2000

2500


Submitted Support requests

Hardware


Hardware

• LCG / AA build servers– 4 linux machines (managed by IT/FIO)– 2 mac osx machines (one ppc)– 2 windows machines

• Special purpose machines– Web / database servers

• Recycling old Hypernews machines (located in CC)

– New machines for• Savannah, Hypernews

Staff & Program of Work


Staff

• Change of project leader– Andreas Pfeiffer left to CMS– Stefan Roiser took over

• 1 staff left, 1 fellow joined– Manuel Gallas left for Atlas– Piotr Kolet joined

• 1 summer student– Luana Micallef

• Currently 3.2 FTE


Program of work

• Completion of the integrated software configuration / testing / release – Web pages derived from LCGCMT– Integrate remaining release steps

• Consistency checking• Production of distribution kits• Documentation generation

– Completion of integration with nightly builds– Integration of installation manager


Program of work (ctd.)

• Nightly build system– Integrate Geant4 builds

• Using IT infrastructure for– Setup of common LCG/AA web services– Setup of database services– Virtual windows machines for s/w build


Program of work (ctd.)

• Hypernews– Investigate usage of IT provided SSO– Find solution for longterm support


Conclusion

• Current focus on LCG/AA s/w stack– Optimize release process– Facilitate and unify processes for

• Nightly builds• Release procedure• User rebuilds

• Outsourcing of maintenance for other services, but still a lot of SPI manpower involved– Savannah– Hypernews– Hardware

SPI Report for the LHCC Comprehensive Review

Documents